Device for simulating a physiological behaviour of a mammal using a virtual mammal, process and computer program

Abstract

A device for simulating a behavior of a mammal in an environment by a virtual mammal includes a mobile head and at least one mobile eye including a first input for successive data representative of eye poses, a second input for mobility action instructions, a memory for storing information on the environment and on stabilization constraints, at least one processor assessing a current part of the environment and recording information on said current part, triggering successive movements of the head and of the eye(s) in function of said mobility action, controlling a dynamic adjustment of the successive movements of the eye(s) with respect to the successive movements of the head in function of the successive data and by using the stabilization constraints. The device further comprises a third input for training data representative of a training movement sequence of the head and the eye(s) associated with a training environment.

The invention claimed is: 
     
       1. A virtual mammal device for simulating a physiological behavior corresponding to a vision system of a virtual mammal in a visual environment, said virtual mammal including: a head mobile with respect to the environment, and at least one eye rotationally mobile with respect to the head, said device comprising:
 at least one first input adapted to receive successive data representative of poses of said at least one eye with respect to the environment; 
 at least one second input adapted to receive instructions for at least one mobility action by the virtual mammal in an environment; 
 at least one memory adapted to store information on the environment and on stabilization constraints between movements of said at least one eye and the head, said stabilization constraints corresponding to at least one of: a vestibulo-ocular reflex, an intentional eye fixation, and an optokinetic reflex; and 
 a computer having at least one processor, the at least one processor of the computer configured to:
 assess a current part of the environment from said successive data, 
 record information on said current part into said at least one memory if said current part is unidentified in said at least one memory, 
 trigger successive movements of the head and said at least one eye as a function of said at least one mobility action, said successive data and the stored information on the environment, and 
 control a dynamic adjustment of the successive movements of said at least one eye with respect to the successive movements of the head as a function of said successive data and the stabilization constraints, 
 
 wherein said device further comprises: at least one third input adapted to receive in a training stage training movement sequence data representative of a training movement sequence of the head and said at least one eye associated with a training environment, and 
 said at least one processor is further configured to learn, during said training stage, at least part of said stabilization constraints through:
 triggering in the training environment successive movements of the head corresponding to said training movement sequence data, 
 assessing a current part of the training environment from said training data, and recording information on said current part into said at least one memory if said current part is unidentified in said at least one memory, 
 determining information on said at least part of the stabilization constraints as a function of said training data, said successive movements of the head, said current part of the training environment and said information on said current part, and 
 recording said information on said at least part of the stabilization constraints into said at least one memory, 
 
 so that the behavior of the virtual mammal executing said at least one mobility action can be simulated, 
 wherein a modular system implemented via the at least one processor includes: functional modules, processing circuitry, ancillary tools, and data structures that are interconnected through a message passing middleware, 
 wherein the virtual mammal device is implemented as a robot via material parts including a structural head, a camera being at least one eye, and effectors that enable movement or activation of the material parts of the virtual mammal device, 
 wherein said at least one processor further comprises at least one first neural network configured to provide said dynamic adjustment through a feedback loop, 
 wherein said at least one processor is further configured to:
 obtain at least one visual stream corresponding to said successive data representative of poses of said at least one eye, assess said current part of the environment from said at least one visual stream, 
 trigger said successive movements, and 
 control said dynamic adjustment as a function of said at least one visual stream, and 
 
 in said training stage,
 obtain at least one training visual stream corresponding to said training movement sequence data representative of the training movement sequence, and 
 assess said current part of the training environment and determine said information as a function of said at least one training visual stream, and 
 
 said device further comprising at least one fourth input adapted to receive at least one simulated visual property, 
 wherein said at least one processor is further configured for dynamically pre-conditioning said at least one visual stream as a function of said at least one simulated visual property, and 
 wherein said at least one simulated visual property is at least partly pertaining to a physiological behavior associated with said at least one eye or to a visual equipment of said at least one eye. 
 
     
     
       2. The device according to  claim 1 , wherein said at least one processor is further configured to:
 simulate at least one working parameter proper to said virtual mammal, and 
 learn, during the training stage, said at least one working parameter in as a function of said training data and of said successive movements of the head, said current part of the training environment and said information on said current part. 
 
     
     
       3. The device according to  claim 1 , wherein said at least one processor is further configured to
 pre-condition said obtained at least one visual stream and said at least one training visual stream, in association with at least one pre-conditioning parameter proper to said mammal to simulate, and 
 learn, during the training stage, said at least one pre-conditioning parameter as a function of said at least one training visual stream, said successive movements, said current part of the training environment and said information on said current part. 
 
     
     
       4. The device according to  claim 1 , wherein said at least one processor is further configured to:
 derive successive images from said at least one visual stream and 
 control said dynamic adjustment by stabilizing at least part of said derived successive images. 
 
     
     
       5. The device according to  claim 1 , wherein the device comprises at least one fifth input adapted to receive information on modified values of the stabilization constraints, said at least one processor is further configured to update the information regarding previous values of the stabilization constraints in the memory with the information regarding said modified values. 
     
     
       6. The device according to  claim 5 , wherein said previous and modified values of the stabilization constraints include at least one of age or a health state of a mammal. 
     
     
       7. The device according to  claim 1 , wherein said at least one processor comprises at least one second neural network configured to assess said current part of the environment and recording the information on said current part. 
     
     
       8. The device according to  claim 1 , wherein said at least one processor is configured to carry out at least one operation including: assessing said current part of the environment, recording said information on said current part, and controlling the dynamic adjustment of the successive movements of said at least one eye, by way of simulated spiking neurons. 
     
     
       9. The device according to  claim 1 , further comprises at least one input adapted to receive information on modified values of the stabilization constraints, said at least one processor being configured for updating the information regarding previous values of the stabilization constraints in the memory with the information regarding said modified values. 
     
     
       10. A method implemented in a virtual mammal for simulating a physiological behavior corresponding to a vision system of a virtual mammal in a visual environment, said virtual mammal being implemented to include: a head rendered mobile with respect to the environment, and at least one eye rendered rotationally mobile with respect to the head, said method comprising:
 receiving successive data representative of poses of said at least one eye with respect to the environment, 
 receiving instructions for at least one mobility action by the virtual mammal in the environment, 
 assessing, by at least one processor, a current part of the environment from said successive data using information on the environment stored in at least one memory, and recording information on said current part into said at least one memory if said current part is unidentified in said at least one memory, 
 triggering, by said at least one processor, successive movements of the head and said at least one eye as a function of said at least one mobility action, said successive data and the stored information on the environment, 
 controlling, by said at least one processor, a dynamic adjustment of the successive movements of said at least one eye with respect to the successive movements of the head as a function of said successive data, and stabilization constraints between movements of said at least one eye and movements of the head, wherein said stabilization constraints corresponding to at least one of: a vestibulo-ocular reflex, an intentional eye fixation, and an optokinetic reflex, said method further comprises: 
 receiving, during a training stage, training data representative of a training movement sequence of the head and said at least one eye associated with a training environment, and 
 learning, by said at least one processor, in said training stage, at least part of said stabilization constraints through:
 triggering in the training environment successive movements of the head corresponding to said training movement sequence, 
 assessing a current part of the training environment from said training data, and recording information on said current part into said at least one memory if said current part is unidentified in said at least one memory, 
 determining information on said at least part of the stabilization constraints as a function of said training data and said successive movements of the head, said current part of the training environment and said information on said current part, and 
 recording said information on said at least part of the stabilization constraints into said at least one memory, 
 
 so that the behavior of the virtual mammal executing said at least one mobility action can be simulated, 
 implementing said at least one processor as further comprising at least one first neural network configured to provide said dynamic adjustment through a feedback loop, and 
 further configuring said at least one processor to:
 obtain at least one visual stream corresponding to said successive data representative of poses of said at least one eye, 
 assess said current part of the environment from said at least one visual stream, 
 trigger said successive movements and control said dynamic adjustment as a function of said at least one visual stream, 
 in said training stage, obtain at least one training visual stream corresponding to said training movement sequence data representative of the training movement sequence, and assess said current part of the training environment and determine said information as a function of said at least one training visual stream, 
 
 implementing at least one input adapted to receive at least one simulated visual property, and said at least one processor being further configured for dynamically pre-conditioning said at least one visual stream as a function of said at least one simulated visual property, wherein said at least one simulated visual property being at least partly pertaining to a physiological behavior associated with said at least one eye or a visual equipment of said at least one eye, 
 wherein a modular system implemented via the processor includes: functional modules, processing circuitry, ancillary tools and data structures that are interconnected through a message passing middleware, and 
 wherein the virtual mammal is implemented as a robot via material parts including: a structural head, a camera being the at least one eye and effectors that enable movement or activation of the material parts of the virtual mammal. 
 
     
     
       11. A non-transitory computer readable medium having stored thereon a computer program comprising instructions which, when the program is executed by a processor, causes said processor to carry out the processes of  claim 10 . 
     
     
       12. The method according to  claim 10 , wherein said at least one processor is further configured to carry out at least one operation including: assessing said current part of the environment, recording said information on said current part, and controlling the dynamic adjustment of the successive movements of said at least one eye, by way of simulated spiking neurons.